THE RESEARCH ON APPLICATION OF FY-2C DATA
IN DROUGHT MONITORING
YU Fan①②
LIU Liang-ming①
WEN Xiong-fei①
①
②
School of Information Engineering and Remote Sensing , Wuhan University , Wuhan
430079,china-yufan021@126.com
Graduate School of the Chinese Academy of Sciences, Beijing 100049, China,-wxfei19@gmail.com
Commission VIII ,WG VIII/2
KEYWORDS: Drought; FY-2C; Cloud Index; Temperature; Drought-monitoring Model;
ABSTRACT:
FY-2C satellite data, with high temporal resolution and vast coverage range, have been applying for long-term and dynamic
drought-monitoring, as well as for weather forecast, especially for forecasting disaster weather. Considering that the traditional
remote sensing drought-monitoring models are easily influenced by cloud, a cloud index-based drought-monitoring model is
presented in this paper. The parameters used in the model are: Max Continuous Cloudy Days (CCD), Max Continuous Cloud-Free
Days (CCFD), Cloudy Days Ratio (CDR) and Temperature Difference from Day to Night (TD). As the important improvement, the
latitude correction function to the three cloud parameters is introduced to the model. Finally, the 20 cm deep soil moisture data from
Oct. 2005 to Sep. 2006 are used to validate the accuracy of this model. The result shows that the improved model is efficacious in the
great extent drought detection and forecast.
1.
2.
INTRODUCTION
MODEL PROFILE
‘Cloud Index-based Drought-monitoring Model' is based on the
following two facts:
Drought is one of the natural calamities in the world; it
influence agricultural produce most in all kinds of the natural
calamities. The average disaster area reaches 20,000,000 km2
for drought in China every year, which makes a loss of 50% of
the whole loss in all of the natural disasters(Liu W T,1996). The
most serious area affected by drought is major grain-producing
region in China. Sustainable drought immediately influences
industrial production, lives of the people and ecological
environment, and even causes various natural calamities, such
as the land desertification, the land subsidence and so on.
Drought has become one of the key restricted factors of the
sustainable
development
of
the
society
and
economy(Gong,1997).
(1) Cloud-free means no precipitation, the possibility of drought
increases. On the contrary, cloud means precipitation, the
possibility of drought decrease.
(2) Cloud-free means receiving more solar short-wave radiation,
the ground temperature rises, and there are more
evapotranspiration, the possibility of drought increases. On the
opposition, cloud can cover the ground, so there is less solar
short-wave radiation, and the result is on the opposition.
2.1 The parameters of the model and its processing
FY-2C is the first operational GEO meteorological satellite of
china successfully launched on Oct. 19 2004 and it was located
at 105°E，FY-2C is in good condition with theirs Visible
Infrared Spin Scan Radiometer (SVISSR), provides the
measurement of the ‘top of atmosphere’ radiance and
reflectance with a temporal resolution of 30 min and a spatial
resolution of 5 km at the sub-satellite point in 4 infrared bands
at 10.8, 12.0, 6.9 and 3.7µm, 1.25 km in 1 visual band at
0.55-0.90µm. Comparing with polar-orbiting satellites, such as
NOAA/AVHRR, FY-2C satellite data has high temporal
resolution, on every half hour to generate an image. As a
geostationary satellite at the same time, the scope of its
coverage is very broad, it is very suitable for Large-scale
drought-monitoring. There are some preprocessing for the
FY-2C data before using it .So a new Drought-monitoring
Model based on FY-2C is intrduced.
2.1.1 Cloud Detection
As the temperature on the top of cloud is much lower than the
land surface in the thermal infrared bands, and the reflectivity of
cloud is extremely high in the visible band, cloud detection can
be completed by simple threshold method with visible bands
and thermal infrared bands. In this paper, the reflectivity
accounted by visible channel and the bright temperature
obtained by an infrared band (10.3 μ m - 11.3 μ m) are used for
cloud detection.
2.1.2
Temperature Difference
Land surface temperature is not only a good indicator of
energy balance of the earth’s surface, but also a key factor of
the physical process in regional or global scale surface. There
are strong correlation between the surface soil moisture and the
land surface temperature difference between day and night. The
greater the temperature difference is, the ability of the surface
evapotranspiration is stronger and the possibility of drought
increases. Due to five channels only in FY-2C, it is complicated
to calculate temperature difference. Bright temperature
509
The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B8. Beijing 2008
difference (TD) is used to replace temperature difference gotten
directly after calibration. Cloud has low bright temperature in
the thermal infrared bands, making it easy for error. So before
the calculation of TD, cloud should be masked. But the TD
computed can not cover the whole research region after cloud
mask. Therefore the average temperature is introduced, which
means the TD over a period of time is used to strike an average,
to make up for the impact of clouds. Then, by the least square
method, the function between TD (average) and the
synchronous drought grade is simulated to get the drought index
of TD (TD_DI).
Cloudy Days (CCD), Max Continuous Cloud-Free Days (CCFD)
and Cloudy Days Ratio (CDR). These three are calculated in a
definite monitoring cycle, the cycle is a month in this research.
The function between CCD and the synchronous drought grade
acquired from the field measurements can be founded by the
least square method. By this function, CCD in the whole image
can be converted to CCD_DI, which is one of the drought-index.
The other two drought-index CCFD_DI and CDR_DI can be
gained in the same way. These three are called drought index of
cloud(Liu,2004).
After drought-index(including CDR_DI, CCFD_DI, CCD_DI,
TD_DI) extracted, the integrated drought index (Int_DI) can be
calculated by the flow chart. (Figure 1).
2.1.3 Cloud-index extraction
Precipitation is an essential factor in the drought monitoring. As
the sole source of precipitation, cloud is very important.
Cloud-index consists of three parameters: Max Continuous
FY-2C raw data
Reprojection and calibration
Cloud detection
TD
Cloud-index
CCFD
CCD
TD_DI
CDR
Cloud_DI
Integrated Drought index(Int_DI)
Figure 1. The flow chart of the model
The integrated drought index can be computed by the
expression as follow:
Where Int_DI is the integrated drought index, X i is the
parameters of drought-index (such as CCFD_DI), and Pi is
Int _ DI = (∑ X i ∗ Pi )
the weight of the drought index.
(1)
In the drought monitoring, the monitoring cycle is a month. The
Int_DI is divided into six categories: very wet D0(-2.0→-1.5)、
wetlands D1(-1.5→-0.5)、normal D2(-0.5→0.5)、slight dry D3
（0.5→1）、dry D4(1→1.5)、heavy drought D5（1.5→2.0）
(Liu Liangming,2005).
510
The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B8. Beijing 2008
3.
MODEL IMPROVEMENTS IN LARGE
SPATIAL SCALES
(astronomical units). In general,
Where
E
The change of latitude leads to diversification of θ and d ,
then makes the solar radiation received by the ground changing.
In order to describe the influence to the drought caused by the
different geographic location of the study area, the introduction
of latitude is necessary in simulating the function between the
cloud and the drought.
In the same region and a certain period of time (such as one
month), the greater the value of CCFD is, the greater the
possibility of drought. Their relationship is positive correlation
and the relations function can be thought as a logarithmic
function( P1 ) approximately, see Figure 2 (a); when it concerns
to CCD, the result is reverse. Bigger the CCD is, the less the
possibility of drought. Their relationship is negative correlation
（2）
is normalized solar radiation,
E0
and an inverse logarithmic function( P2 ) can be used to
describe their relation, see Figure 2 (b); Also the greater the
value of CDR is, the smaller the possibility of drought, and their
is normalized
relationship can be depicted as a linear function ( P3 ), see
solar radiation at the average distance between the earth and
sun , θ is solar zenithal angle(complementary angle of solar
altitude angle),
d
Figure 2 (c). In the Figure 2, the longitudinal coordinates DI is a
single parameter of integrated drought index (Int_DI), range (-2,
+2); the abscissas are respectively corresponding to the value of
CCFD, CCD and the CDR in a cycle of 30 days, range(0,30).
is the distance between the earth and sun
(a)
is a constant , θ can be
obtained from the raw data, d is a season-related variable。
In the large-scale drought-monitoring, even the same drought
conditions leads to different results because of the different
research areas. For example, within one month, if Max
Continuous Cloud-Free Days (CCFD) is 10 in the South of
China, it is probably very dry. But towards the north of China,
may be the situation is not serious. The different geographical
position of the study area influences the result of drought
monitoring. The reason mainly is that the ground of different
regions receives different solar irradiance (or the total solar
radiation energy). In Formula 2, the solar altitude angle and the
distance from the sun to the earth cause the diversity.
E cos θ
E= 0 2
d
E0
(b)
(c)
Figure 2. the relation function between DI and the three cloud parameters in the same area
As derived from the research on the small and restricted area,
the functions in Figure 2 can not be used in the large area
research directly. The introduction of latitude to amend cloud
parameters in different area is feasible. when latitude increases
in different regions, the sun altitude angle decreases, then the
solar radiation energy received by the ground
decreases(formula 2). In order to achieve the same Int_DI, the
value of CCFD should become bigger, the correction function
of CCFD can be describe as monotone increasing function
(e.g., exponential function), shown in Figure 3 (a); In the same
way, the amend function of CCD can be described as
monotone decreasing function, and also monotone decreasing
function can describe the correction function of CDR, see
Figure 3 (b), Figure 3 (c). In the Figure 3(a), Figure 3 (b) and
Figure 3 (c), the longitudinal coordinates is latitude, the
abscissas are respectively corresponding to CCFD，CCD and
CDR.
511
The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B8. Beijing 2008
(b)
(a)
(c)
Figure 3. three correction functions
4. CASE ANALYSIS
The research after large number of experiments shows that: the
cosine of latitude can be better fitted for the three correction
functions, which can be seen in Eq 3:
The research zone focus on the Chinese land area, latitude is
from north 18 ° 05 'to north 53 ° 34' , longitude is form east
longitude 73 ° 20 'to east 135 ° 05'. The data is the FY-2C data
from October 2005 to September 2006.
CCFD ' = P1 (CCFD ∗ S1 (lat ))
CCD ' = P2 (CCD ∗ S 2 (lat ))
Analysis of the relationship between the parameters and the
drought index can determine the relation function. Many
experiments show that the relation function between CCD and
Int_DI or CCFD and Int_DI is nearly exponential function. The
function of CDR or CFDR can be used with the sub-linear
function; the function between temperature difference and
Int_DI can be described as almost exponential function[11].
Some of the functions are in the Table 1.
(3)
CDR ' = P3 (CDR ∗ S 3 (lat ))
CCFD ' , CCD ' and CDR ' are the cloud
parameters after correction; CCFD , CCD and CDR
are the cloud parameters before correction， lat is latitude.
Where
CCFD
Drought-index
CCD
Drought-index
CDR
Drought-index
TD
Drought-index
1
2
3
4
5
6
7
8
9
10
≥11
-2
-1.8
-1.5
-1.2
-0.5
0.2
0.8
1.2
1.7
2
1
2
3
4
5
-0.2
-0 2
6
7
8
9
10
≥11
2
1.8
1.2
0.8
0
-0.5
-1.2
-1.2
-1.6
-1.8
-2
≤20
（20 ～ 50）
（50 ～ 70）
2
-（X-40）/10
-（X-50）/20 -1
（3 ～ 7）
（7 ～ 11）
≤3
-2
2
2
-（7-X） /8
（X-7） /8
≥70
-2
≥11
2
Table 1 Functions of all parameters from June to August
4.1 Qualitative analysis
Comparing the model result(see Appendix) with the
synchronous Distribution of Drought Climate (China national
drought monitoring business products, in 1995 R & D) from Oct.
2005 to Sep. 2006, the distribution of drought are basically the
same. For example, the distribution maps shows severe drought
512
on January, February and May in Yunnan province and terrible
long-term drought in Chongqing from July to September. The
outcome is proved correct by the Distribution of Drought
Climate. Taking some maps for example, the contrasts are as
follows (Figure 4, Figure 5):
The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B8. Beijing 2008
(a)
(b)
Figure 4. The outcome of model and the Distribution of Drought Climate on Mar. 2006
(a)
(b)
Figure 5. The outcome of model and the Distribution of Drought Climate on Apr. 2006
There are some differences in the northwest of China in the
Figure 4 and Figure 5. Because the northwest of China are
desert and hungriness, undoubtedly they are very arid, and there
is little field observation station there, so the Distribution of
Drought Climate based on field observation shows no drought.
Droughts of the other place are nearly the same in both of the
maps.
4.2 Quantitative analysis
The verification data provided by the National Satellite
Meteorological Center (NSMC) is 20 cm deep soil moisture
from Oct. 2005 to Sep. 2006 obtained by 810 field stations in
China. The relationship between 20 cm deep soil moisture
percentage and drought-grade can be found on the
Agro-meteorological Observing Criterion(Qin,2003), the
standards is shown in Table 2:
very wet D0
wetlands D1
normal D2
little dry D3
dry D4
heavy drought D5
94～99%
80～93%
61～79%
51～60%
41～50%
≦40%
Table2 The relationship between 20 cm deep soil moisture percentage and drought-grade
So the precision can be quantitatively calculated by the droughty grade obtained by Eq (1) minus the Verification grade gotten by
Table5, the distribution table is below (Table 3):
Year
Month
Precision
homology
One
grade
discrepancy
Two
grade
discrepancy
Three
grade
discrepancy
>3
grade
discrepancy
2005 年
Orc
Nov
2006 年
Dec
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
34%
39%
37%
38%
33%
41%
28%
38%
30%
40%
33%
38%
37%
37%
38%
37%
39%
41%
40%
43%
38%
39%
44%
37%
20%
15%
11%
19%
15%
17%
19%
18%
11%
9%
16%
12%
6%
6%
10%
12%
10%
9%
6%
7%
9%
8%
5%
7%
1%
4%
5%
3%
5%
3%
1%
0%
0%
0%
2%
0%
Table 3 The precision of model from Oct. 2005 to Sep. 2006
513
The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B8. Beijing 2008
Statistics from the accuracy table of the results of the drought
monitoring shows that the precision of the grade discrepancy of
zero and one can control in 66% -83%. With the exception of
66% on Jan. 2006, accuracy of the other month is more than
70%. The error of two grade discrepancy does not exceed 20%
and the three grade discrepancy error and above is no more than
15%. As the study area is the whole inland of China, many
factors influence the precision, such as anomalistic topography,
intricate climate, different season and so on. Also there are
some meteorological stations in the especial location and they
can not reflect the drought. The Precision is acceptable because
the precision of the grade discrepancy of zero and one can
control nearly over 70%. The precision result shows that the
model is reliable and stable.
4. There are many influencing factors in the large-scale drought
monitoring, such as artificial rainfall, artificial irrigation, human
destruction of the ecological environment, and other activities,
which may aggravate or relieve the drought. The difficulty is
these factors are unable to control and quantitative analysis.
REFERENCE
[1] Carlson，T . N.， Regional scale estimates of surface
moisture availability and thermal inertia using remote thermal
measurements ［J］. Remote Sensing Reviews，1986（1）：
197 ～ 247 .
[2] Gong Deji, Hao Muling, Hou Qiong. Study on the
Complex Index of the Drought Disaster. Meteorological
Monthly,1997 ,22 (10) :3～7
5. CONCLUSION
This is the first time to monitor drought in the whole Chinese
inland, the problem of how to monitor drought in the large-scale
area and the adaptability of the model has been solved. By the
research, the three correction function are set up (Eq (3)), which
was proved to be efficient and reliable. From the precision
distribution in the Table 6, the Cloud Index-based
Drought-monitoring Model can basically meet the application
needs.
[3] Liu W T，Kogan F．Monitoring regional drought using the
vegetation condition index[J]．Int J Remote Sense，1996，1
7(14)：2761～2782．
[4] Liu Liangming，The Research of Remote Sensing Drought
Prediction Model Based on EOS MODIS Data ， A Ph.D.
Dissertation of Wuhan University, P.R. China. 20004
But there are some shortcomings in the model as follows:
1. The spatial resolution of FY-2Cdata is 5000m*5000m, a
pixel represents an area of 25km2, the verification data from the
field station denotes a very small area, nearly a point. It is not
very suitable to validate the result of the model by this
verification data.
2. The temperature of cloud top is close to the land surface
temperature in the place of high latitude in winter. Drought
detecting by the threshold method may leads error of the
correction function.
3. The precision of the Cloud Index-based Drought-monitoring
Model depends on the integrity and continuity of FY-2C. The
incomplete FY-2C data can not get the real CCDF and CCD,
and can not reflect the real situation of drought.
[5] Liu Liangming，Hu Yan. Analysis of Parameters and
Their Powers of MODIS Drought Monitoring Model.
Geomatics and Information Science of Wuhan University,
2005，30（2）：139～142
[6]
Qin Dahe.e.t. Drought[M].weather book concern.2003.
ACKNOWLEDGEMENTS
We thank the National Satellite Meteorological Center for
supplying the FY-2C data.
514
The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B8. Beijing 2008
APPENDIX
The distribution maps of drought from Oct. 2005 to Sep. 2006
515
The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B8. Beijing 2008
516